Phase-Type Fitting Using HyperStar

摘要

In this paper we provide a hands-on discussion of the use of the HyperStar phase-type fitting tool in common application scenarios. HyperStar allows fitting Hyper-Erlang distributions to empirical data, using a variety of algorithms and operation modes. We describe simple cluster-based fitting, a new graphical method for refining the density approximation, a new command-line interface, and the integration of HyperStar with a Mathematica implementation of a fitting algorithm. Furthermore, we describe the use of Hyper-Erlang distributions in simulation. Throughout our discussion we illustrate the concepts on a data set which has been shown to be difficult to fit with a PH distribution.